High-precision building surveying and mapping method and system

Abstract

The application relates to the technical field of building surveying and mapping, in particular to a high-precision building surveying and mapping method and system. The method comprises the following steps: collecting a point cloud model and an image of a building to be surveyed and mapped, segmenting the point cloud model into point cloud regions, obtaining difference coefficients of the point cloud regions, identifying building regions and removing point cloud data outside the building regions, and performing hole filling on the removed point cloud model; obtaining structure surfaces of the building regions, obtaining discrete coefficients of boundary points and non-boundary points of the structure surfaces, and screening noise suspected points from all the non-boundary points of the structure surfaces; obtaining reliable coefficients and point cloud distortion coefficients of the building regions, performing noise removal processing on the point cloud data of the building regions, and obtaining a three-dimensional model of the building to be surveyed and mapped. The application aims to improve the noise removal precision of the point cloud data and the precision of the three-dimensional model, thereby providing more accurate and more intuitive data support for subsequent building surveying and mapping and engineering cost assessment.

Description

Technical Field

[0001] This application relates to the field of architectural surveying technology, specifically to a high-precision architectural surveying method and system. Background Technology

[0002] With continuous innovation in architectural design, the geometric shapes of irregularly shaped buildings have brought new challenges to as-built surveying and engineering cost assessment. However, with the development of technologies such as UAV aerial surveying and 3D laser scanning, architectural surveying is still moving towards higher precision and efficiency. High-precision architectural surveying helps to obtain complete and accurate building information, providing a reliable data foundation for BIM modeling and building quantity calculation, thereby improving the accuracy and scientific nature of engineering cost assessment.

[0003] Existing technologies typically utilize 3D laser scanners to acquire point cloud data of complex buildings and then denoise the acquired point cloud data to improve the accuracy of subsequent building model construction, surveying, and cost assessment. However, existing technologies usually perform overall denoising based on the overall noise level of the point cloud data, without fully considering the variations in noise characteristics in different areas of a complex building due to differences in materials or structures. If uniform parameters are used for processing, key structural points in complex buildings may be misidentified as noise and removed, thus affecting the surveying accuracy of complex buildings and the scientific validity of cost assessments. Summary of the Invention

[0004] In view of the above, it is necessary to provide a high-precision building surveying method and system. Compared with traditional high-precision building surveying methods and systems, this method improves the denoising accuracy of point cloud data and the accuracy of 3D models, thereby providing more accurate and intuitive data support for subsequent building surveying and engineering cost assessment.

[0005] In a first aspect, embodiments of this application provide a high-precision building surveying method, which includes the following steps:

[0006] Collect point cloud models and images of the building to be surveyed. The point cloud model contains the color and reflection intensity of the point cloud data.

[0007] The point cloud model is divided into point cloud regions by point cloud segmentation method. The difference coefficient of each point cloud region is obtained by the spatial overlap, color difference and reflection intensity difference of point cloud data between each point cloud region and other point cloud regions. Then, each building region is identified and the point cloud data outside the building region is removed from the point cloud model. The hole is filled in the removed point cloud model.

[0008] The structural surfaces of each building area are segmented to obtain the structural surfaces and their interior point ratios. Boundary points and non-boundary points of each structural surface are also obtained. The dispersion of reflection intensity of boundary points and non-boundary points within a preset nearest neighbor range is used to obtain the dispersion coefficients of each boundary point and non-boundary point. This allows for the selection of suspected noise points from all non-boundary points on each structural surface. The reliability coefficient of each building area is obtained by combining the interior point ratio of each structural surface with the distribution of the dispersion coefficients of all boundary points. Suspected noise points in each structural surface are classified according to reflection intensity. The point cloud distortion coefficient of each building area is obtained by combining the color and curvature dispersion of suspected noise points in each category within each building area, the normal vector similarity between suspected noise points in each structural surface, and the reliability coefficient. A statistical filtering algorithm is then used to denoise the point cloud data of each building area. Finally, combined with the image, a 3D model of the building to be surveyed is obtained.

[0009] In one embodiment, the process of obtaining the difference coefficient is as follows:

[0010] Obtain the AABB bounding box of each point cloud region, and obtain the overlap value of the AABB bounding box between each point cloud region and each other point cloud region on the three coordinate axes; denot each point cloud region whose overlap value with each other on the three coordinate axes is positive as each connected point cloud region of each point cloud region.

[0011] Calculate the magnitude of the vector composed of the R, G, and B color components of each point cloud data in each point cloud region; calculate the arithmetic mean of the magnitudes of all point cloud data in each point cloud region; and calculate the average value of the reflection intensity of all point cloud data in each point cloud region.

[0012] The average of the differences between the arithmetic mean of each point cloud region and all its connected point cloud regions is denoted as the first mean; the average of the differences between the average of each point cloud region and all its connected point cloud regions is denoted as the second mean.

[0013] The difference coefficients are directly proportional to the first mean and the second mean, respectively, and inversely proportional to the number of connected point cloud regions in each point cloud region.

[0014] In one embodiment, the difference coefficient is calculated as follows: the sum of the first mean and the second mean is calculated, the quantity is mapped to a positive number, and the difference coefficient is the ratio of the sum to the positive number.

[0015] In one embodiment, the method for identifying each building area and removing point cloud data outside the building areas from the point cloud model is as follows:

[0016] Obtain the segmentation threshold of the difference coefficient of all point cloud regions, and regard the point cloud regions with the difference coefficient less than or equal to the segmentation threshold as building regions, and remove the point cloud data outside the building regions from the point cloud model.

[0017] In one embodiment, the method for obtaining the discrete coefficients is as follows:

[0018] The ratio of the dispersion of the reflection intensity of all boundary points within the preset nearest neighbor range of each boundary point of each structural surface to the number of boundary points is used as the dispersion coefficient of each boundary point of each structural surface.

[0019] The ratio of the dispersion of the reflection intensity of all non-boundary points within a preset nearest neighbor range of each non-boundary point on each structural surface to the number of non-boundary points is used as the dispersion coefficient of each non-boundary point on each structural surface.

[0020] In one embodiment, the process of obtaining the suspected noise point is as follows:

[0021] An anomaly detection algorithm is used to obtain the anomaly score of the discrete coefficients of each non-boundary point of each structural surface. Boundary points whose normalized anomaly score is greater than a preset threshold are identified as suspected noise points.

[0022] In one embodiment, the reliability coefficient is obtained by:

[0023] An anomaly detection algorithm is used to obtain the anomaly score of the discrete coefficient of each boundary point of each structural surface, and the proportion of boundary points with normalized anomaly scores greater than a preset threshold in each structural surface is counted.

[0024] The reliability coefficient is directly proportional to the average inward point ratio of all structural surfaces of each building surface, and inversely proportional to the average proportion of the number of structural surfaces in each building surface.

[0025] In one embodiment, the process of obtaining the point cloud distortion coefficients is as follows:

[0026] Calculate the normal vector of each suspected noise point, and use the region growing method to obtain each growing region in each structural surface based on the similarity of the normal vectors between suspected noise points in each structural surface. Calculate the proportion of suspected noise points outside the growing region in each structural surface among all suspected noise points.

[0027] The dispersion of the modulus of all suspected noise points in each class is denoted as the color dispersion; the mean of the color dispersion and the curvature dispersion in each class is taken as the discrete mean of each class.

[0028] The expression for the point cloud distortion coefficient of each building area is as follows:

[0029] In the formula, This represents the point cloud distortion coefficient for the u-th building region; This represents the average of the discrete means of all classes across all structural surfaces in the u-th building region; This represents the average percentage of the number of such structural surfaces in all structural surfaces of the u-th building region; This represents the reliability coefficient of the u-th building area.

[0030] In one embodiment, during the process of denoising the point cloud data of each building area using a statistical filtering algorithm, the negative of the point cloud distortion coefficient of each building area is used as the exponent of an exponential function with the natural constant as the base, and the calculation result of the exponential function is used as the standard deviation multiple when the statistical filtering algorithm denoises the point cloud data of each building area.

[0031] Secondly, embodiments of this application also provide a high-precision building surveying system, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the above-described high-precision building surveying methods.

[0032] This application has at least the following beneficial effects:

[0033] This application obtains a difference coefficient by measuring the spatial overlap, color difference, and reflectivity difference between point cloud regions. This coefficient quantifies the difference between each point cloud region and its surrounding region, which helps in the subsequent identification of occluding objects that are not part of the building itself, thus improving the realism and usability of the point cloud data. By identifying building regions and removing point cloud data outside of those regions, the computational load of subsequent processing is reduced. Filling holes in the removed point cloud model ensures the integrity of the point cloud model.

[0034] Furthermore, by obtaining reliability coefficients, the degree of interference to the point cloud data of each building area can be assessed, providing dynamic parameter basis for subsequent denoising processing. By filtering suspected noise points using discrete coefficients and analyzing the color and curvature consistency of these points under the condition of consistent reflection intensity, combined with the spatial continuity and concentration characteristics of suspected noise points, point cloud distortion coefficients are obtained. This quantifies the actual degree of interference to the point cloud data of each building area, providing more accurate parameter basis for subsequent denoising processing. Dynamically adjusting the parameters in the statistical filtering algorithm based on the point cloud distortion coefficients enables targeted denoising according to the noise characteristics of different areas, preserving key architectural details, improving the accuracy of the 3D model, and combining images to apply textures to the 3D model, generating a textured 3D model. This provides more accurate and intuitive data support for subsequent architectural surveying and engineering cost assessment. Attached Figure Description

[0035] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 A flowchart illustrating the steps of a high-precision building surveying method provided in one embodiment of this application;

[0037] Figure 2 This is a schematic diagram illustrating the process of obtaining the difference coefficient.

[0038] Figure 3 This is a schematic diagram illustrating the process of obtaining the point cloud distortion coefficient. Detailed Implementation

[0039] In the description of the embodiments in this application, the words "exemplary," "or," and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary," "or," and "for example" is intended to present the relevant concepts in a specific manner.

[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. It should be understood that, unless otherwise stated, " / " in this application means "or".

[0041] It should also be noted that the terms "first" and "second" in this application are used to distinguish similar objects, rather than to describe a specific order or sequence.

[0042] The following description, in conjunction with the accompanying drawings, details the specific scheme of the high-precision architectural surveying method and system provided in this application.

[0043] Please see Figure 1 The diagram illustrates a flowchart of a high-precision building surveying method according to an embodiment of this application. The method includes the following steps:

[0044] Step 1: Collect the point cloud model and image of the building to be surveyed. The point cloud model contains the color and reflection intensity of the point cloud data.

[0045] Any irregularly shaped building with a complex structure is taken as the building to be surveyed. Four survey stations are set up in the four directions (front, back, left, and right) of the building, with a distance of 50 meters between each station and the building. Three targets are placed between adjacent stations. A 3D laser scanner with color scanning capability is used to scan the building at each station, with a scanning interval of 15 mm. After scanning, the point cloud data collected from adjacent stations are stitched together to obtain a point cloud model of the building. The point cloud model contains the color and reflection intensity of the point cloud data. It should be noted that 4, 50 meters, and 3 are merely one embodiment of this application. Implementers can set specific values ​​according to actual conditions, and this application does not impose any special restrictions.

[0046] In this embodiment, the ICP (Iterative Closest Point) algorithm is used to stitch point cloud data. The ICP algorithm is a well-known technology and will not be described in detail in this application. As other implementation methods, based on the ability to stitch point cloud data, implementers may use other existing technologies, such as the FPFH (Fast Point Feature Histograms) algorithm, etc. This application does not impose any special restrictions.

[0047] In addition, drones equipped with high-definition cameras are used to fly around the building to be surveyed, capturing high-definition images of the building from multiple angles, thereby providing data support for the subsequent texture mapping of the 3D model of the building.

[0048] Step 2: The point cloud model is divided into point cloud regions using a point cloud segmentation method. The difference coefficient of each point cloud region is obtained by considering the spatial overlap, color difference, and reflection intensity difference between the point cloud data of each region and the other point cloud regions.

[0049] When acquiring point cloud data of a building to be surveyed, obstructions such as trees, vehicles, and pedestrians often exist around the building. These interfering objects are scanned into the original point cloud data, resulting in redundant information or geometric errors in the generated point cloud model. Therefore, it is necessary to identify and remove obstructing objects other than the building itself to improve the realism and usability of the point cloud data, thereby improving the efficiency and accuracy of subsequent building surveying and engineering cost assessment.

[0050] The point cloud model of the building to be surveyed is segmented using the point cloud segmentation method to obtain each point cloud region. Taking the i-th point cloud region as an example, the confidence level that the i-th point cloud region is a non-building body is analyzed.

[0051] In this embodiment, a PointNet network trained on the point cloud segmentation dataset Semantic3D is used to segment the point cloud model of the building to be surveyed. The point cloud segmentation dataset Semantic3D is a well-known dataset and will not be described in detail here. As other implementation methods, based on the ability to segment the point cloud model of the building to be surveyed, the implementer may use other existing feasible technologies, such as segmentation algorithms based on region growing, etc. This application does not impose any special restrictions.

[0052] Because most obstructions are spatially isolated, they have gaps between themselves and the main building, rather than being continuous or closely connected like building components.

[0053] Based on the above analysis, the maximum values ​​of the point cloud data in the i-th point cloud region on the X, Y, and Z coordinate axes are obtained respectively. and minimum value This is used to obtain the AABB bounding box of the i-th point cloud region using the AABB bounding box algorithm. The AABB bounding box algorithm is a well-known technique and will not be described further in this application. Following the method used to obtain the AABB bounding box of the i-th point cloud region, the AABB bounding boxes of the remaining point cloud regions are obtained.

[0054] The AABB bounding box of the i-th point cloud region is sequentially combined with the AABB bounding boxes of all other point cloud regions as input to the calOverTotal method for AABB bounding box collision detection. The overlap values ​​of the AABB bounding boxes of the i-th point cloud region with each of the other point cloud regions on the X, Y, and Z axes are output sequentially. Each point cloud region with a positive overlap value on all three axes with the i-th point cloud region is denoted as a connected point cloud region of the i-th point cloud region. The calOverTotal method is a well-known technique and will not be elaborated upon here. The number of connected point cloud regions of the i-th point cloud region reflects the number of objects that overlap with the i-th point cloud region in space. The smaller the number, the more isolated the i-th point cloud region is, and the greater the probability that the i-th point cloud region is an occluder.

[0055] Furthermore, considering the possibility of special cases where obstructions connect to buildings, such as pedestrians opening building doors or vehicles hugging building walls, further analysis and processing are required. Since obstructions and the obscured objects typically differ in color and material, the point cloud color and reflection intensity also exhibit significant abrupt changes.

[0056] Calculate the magnitude of the vector composed of the R, G, and B color components of each point cloud data within each point cloud region; calculate the arithmetic mean of the magnitudes of all point cloud data within each point cloud region; calculate the average of the reflection intensity of all point cloud data within each point cloud region; denot the average difference between the arithmetic mean of the i-th point cloud region and all its connected point cloud regions as the first mean; denot the average difference between the average values ​​of the i-th point cloud region and all its connected point cloud regions as the second mean. The first mean and the second mean respectively reflect whether there are significant color and material differences between the i-th point cloud region and its connected point cloud regions.

[0057] In this embodiment, the difference value is the absolute value of the difference. As other implementation methods, based on the ability to measure the degree of difference between the arithmetic mean and the average value, the implementer may use other calculation methods, such as the square of the difference, etc. This application does not impose any special restrictions.

[0058] Furthermore, the difference coefficient of the i-th point cloud region is obtained by using the first and second means of the i-th point cloud region and the number of connected point cloud regions of the i-th point cloud region. The expression is as follows:

[0059] ; This represents the difference coefficient of the i-th point cloud region; This represents the sum of the first mean and the second mean of the i-th point cloud region; This represents the number of connected point cloud regions in the i-th point cloud region; τ represents a preset positive number to avoid a denominator of 0. The value of τ is preset by the user and can be set by the implementer. In this embodiment, the value of τ is 0.01. If there is a case where the number of connected point cloud regions in the i-th point cloud region is 0, in order to ensure the normal calculation of the formula, ... The value is set to 1.

[0060] It should be noted that the difference coefficient reflects the probability that the i-th point cloud region is an occluding object around the building to be mapped. The smaller the number of connected point cloud regions of the i-th point cloud region, and the greater the difference in color and material between the i-th point cloud region and its connected point cloud regions, the larger the difference coefficient, indicating that the i-th point cloud region is more likely to be an occluding object.

[0061] Following the same method used to obtain the difference coefficient of the i-th point cloud region, the difference coefficients of the remaining point cloud regions are obtained. A schematic diagram of the difference coefficient acquisition process is shown below. Figure 2 As shown.

[0062] Step 3: Identify each building region and remove point cloud data outside the building region from the point cloud model; fill holes in the removed point cloud model; segment each building region into structural surfaces, obtain each structural surface and its interior point ratio, obtain each boundary point and non-boundary point of each structural surface; obtain the dispersion coefficient of each boundary point and non-boundary point of each structural surface by using the dispersion of the reflection intensity of each boundary point and non-boundary point within the preset nearest neighbor range of each structural surface, combined with the number of boundary points and non-boundary points within the preset nearest neighbor range; obtain the reliability coefficient of each building region by using the interior point ratio of each structural surface of each building region, combined with the distribution of the dispersion coefficients of all boundary points of each structural surface.

[0063] Furthermore, since the building to be surveyed is assembled from multiple areas, and the shape complexity and surface texture characteristics of different areas vary, the degree of interference experienced by different areas is also inconsistent. Therefore, it is necessary to analyze the degree of interference experienced by each area in order to more accurately eliminate the influence of interference and improve the accuracy of building surveying and cost assessment.

[0064] The segmentation threshold of the difference coefficient of all point cloud regions in the point cloud model is obtained. Point cloud regions with difference coefficients less than or equal to the segmentation threshold are regarded as building regions. Point cloud data outside the building regions in the point cloud model are removed, thereby reducing the amount of data in subsequent point cloud model processing and improving processing efficiency and accuracy.

[0065] In this embodiment, the Otsu threshold segmentation algorithm is used to obtain the segmentation threshold of the difference coefficient of all point cloud regions of the point cloud model. The Otsu threshold segmentation algorithm is a well-known technology and will not be described in detail in this application. As other implementation methods, based on the ability to obtain the segmentation threshold of the difference coefficient of all point cloud regions of the point cloud model, the implementer may use other existing technologies, such as iterative threshold segmentation, global threshold segmentation, etc. This application does not impose any special restrictions.

[0066] After removing non-building structures, considering that the building areas connected to occlusions may have holes or missing parts, the holes in the point cloud model formed by the point cloud data outside the building areas are filled.

[0067] In this embodiment, a scattered point cloud model repair method is used to fill the holes in the point cloud model. The scattered point cloud model repair method is a well-known technology and will not be described in detail in this application. As other implementation methods, based on the ability to fill the holes in the point cloud model, implementers can use other existing feasible technologies, and this application does not impose any special restrictions.

[0068] Structural surfaces are segmented for each building area to obtain each structural surface and the interior point ratio of each structural surface. Boundary points of each structural surface are also obtained. The point cloud data of each structural surface, excluding the boundary points, are recorded as non-boundary points.

[0069] In this embodiment, the RANSAC (Random Sample Consensus) algorithm is used to obtain the structural surfaces of each building area and the interior point ratio of each structural surface. The plane with the most interior points obtained after a preset number of iterations is taken as a structural surface. Interior points belonging to the structural surface are removed from the point cloud data of each building area. This process of obtaining structural surfaces is repeated until the remaining number of points is less than a preset threshold. The interior point ratio of each structural surface is the ratio of the number of interior points belonging to each structural surface to the total number of point cloud data involved in obtaining each structural surface. For example, if there are 1000 point cloud data points, and the RANSAC algorithm is used to fit a structural surface, and 800 of the 1000 point cloud data points are interior points of the structural surface, then the interior point ratio of the structural surface is 0.8. The preset number is 1000, and the preset threshold is 200. The values ​​of the preset number and the preset threshold are preset manually, and the implementer can set them according to the actual situation. This application does not impose any special restrictions. The RANSAC algorithm is a known technology, and this application does not impose any special restrictions on it. As an alternative implementation, based on the availability of structural surfaces and internal point ratios for each building area, implementers may employ other existing technologies, such as the RegionGrowing class of the PCL point cloud library, etc. This application does not impose any special restrictions.

[0070] In this embodiment, the point cloud data of each structural surface is used as the input of the AC (Angle Criterion) method in the PCL point cloud library, and the output is each boundary point of each structural surface. The AC method in the PCL point cloud library is a well-known technology, and will not be described in detail in this application. As other implementation methods, based on the ability to obtain each boundary point of each structural surface, the implementer may adopt other existing feasible technologies, and this application does not impose any special restrictions.

[0071] Taking the z-th structural surface of the u-th building region as an example, the inlier rate of the z-th structural surface reflects the structural regularity and perturbation degree of the point cloud data of the z-th structural surface. The higher the inlier rate, the better the fitting effect of the z-th structural surface and the less noise interference it is affected by.

[0072] Furthermore, since noise points are usually isolated points with few surrounding points, while true boundary points are more concentrated, with a large number of points in the neighborhood, and the reflection intensity of true boundary points is more consistent.

[0073] Based on the above analysis, taking the w-th boundary point of the z-th structural surface as the center, each boundary point with an Euclidean distance less than q from the w-th boundary point is considered its nearest neighbor. The dispersion of the reflection intensity of all nearest neighbors of the w-th boundary point is calculated, and the ratio of the dispersion to the number of nearest neighbors of the w-th boundary point is taken as the dispersion coefficient of the w-th boundary point. The dispersion coefficient reflects the consistency of the point cloud reflection intensity and the degree of point cloud aggregation at the w-th boundary point. The smaller the dispersion coefficient of the w-th boundary point, the more likely the w-th boundary point is to be a true boundary point.

[0074] In this embodiment, the value of q is 10. The value of q is preset by the user and can be set by the implementer according to the actual situation. This application does not impose any special restrictions.

[0075] In this embodiment, the dispersion of the reflection intensity is the variance. As other implementation methods, based on the ability to measure the uneven distribution of the reflection intensity of neighboring points, implementers may use other existing technologies, such as standard deviation, coefficient of variation, etc. This application does not impose any special restrictions.

[0076] The discrete coefficients of each boundary point of each structural surface are obtained by following the method for obtaining the discrete coefficients of the w-th boundary point of the z-th structural surface.

[0077] An anomaly detection algorithm is used to obtain the anomaly scores of the discrete coefficients of each boundary point of the z-th structural surface. The proportion of boundary points in the z-th structural surface whose normalized anomaly scores are greater than a preset threshold is counted. The proportion reflects the proportion of boundary points affected by anomaly interference in the boundary region of the z-th structural surface. The smaller the proportion, the less interference the z-th structural surface is affected.

[0078] In this embodiment, the LOF (Local Outlier Factor) anomaly detection algorithm is used to obtain the anomaly score of the discrete coefficients of each boundary point of the z-th structural surface. The LOF value is used as the anomaly score. The LOF anomaly detection algorithm is a well-known technology and will not be described in detail in this application. As other implementation methods, based on the ability to measure the anomaly degree of the discrete coefficients of each boundary point of the z-th structural surface, the implementer may use other existing technologies, such as the isolated forest algorithm, etc. This application does not impose any special restrictions.

[0079] In this embodiment, the hyperbolic tangent function is used to obtain the normalized value of the abnormal score of the discrete coefficient of the boundary point. The hyperbolic tangent function is a well-known technique and will not be described in detail in this application.

[0080] In this embodiment, all preset threshold values ​​are 0.5. The preset threshold values ​​are preset by humans, and implementers can set them according to actual conditions. This application does not impose any special restrictions.

[0081] Furthermore, by combining the interior point ratio of each structural surface in the u-th building region with the aforementioned quantity ratio in each structural surface of the u-th building region, the reliability coefficient of the u-th building region is obtained, expressed as:

[0082] In the formula, This represents the reliability coefficient of the u-th building area; This represents the mean of the interior point ratios of all structural surfaces in the u-th building region; This represents the average percentage of the number of such structural surfaces in all structural surfaces of the u-th building region; This indicates a preset positive number, used to avoid a denominator of 0. The value is preset by a person, and the implementer can set it himself. In this embodiment... The value is 0.01.

[0083] It should be noted that the reliability coefficient reflects the degree of interference to the point cloud model of the u-th building area. The smaller the reliability coefficient, the greater the interference when laser scanning is performed on the u-th building area in the building to be surveyed, and the more effort is needed to remove the interference.

[0084] The reliability coefficients of the remaining building areas are obtained using the same method as for obtaining the reliability coefficient of the u-th building area.

[0085] Step 4: Screen all suspected noise points from all non-boundary points of each structural surface; classify the suspected noise points in each structural surface according to reflection intensity; obtain the point cloud distortion coefficient of each building area by combining the color dispersion and curvature dispersion of the suspected noise points in each category in each building area, as well as the normal vector similarity between the suspected noise points in each structural surface of each building area, and the reliability coefficient.

[0086] Considering that the building to be surveyed may contain irregular objects such as sculptures and sloping, irregularly shaped walls; and that to highlight the artistry of irregular structures, the design often incorporates variations in materials and colors to enhance visual prominence, resulting in significant differences in the reflection intensity of point clouds between critical and non-critical areas. Therefore, when analyzing irregular objects, decorative elements or normal geometric changes may be misjudged as abnormal interference, leading to errors in the reliability coefficient, requiring further analysis.

[0087] Following the method for obtaining the discrete coefficients of the w-th boundary point of the z-th structural surface, the discrete coefficients of each non-boundary point of each structural surface are obtained. When calculating the discrete coefficients of non-boundary points, it is only necessary to replace the boundary points involved in the calculation of the discrete coefficients of the boundary points with non-boundary points.

[0088] An anomaly detection algorithm is used to obtain the anomaly score of the discrete coefficients of each non-boundary point of each structural surface. Boundary points whose normalized anomaly score is greater than a preset threshold are identified as suspected noise points.

[0089] In this embodiment, the LOF (Local Outlier Factor) anomaly detection algorithm is used to obtain the anomaly score of the discrete coefficients of each non-boundary point of each structural surface. The LOF value is used as the anomaly score. The LOF anomaly detection algorithm is a well-known technology and will not be described in detail in this application. As other implementation methods, based on the ability to measure the degree of anomaly of the discrete coefficients of each non-boundary point of each structural surface, the implementer may use other existing technologies, such as the isolated forest algorithm, etc. This application does not impose any special restrictions.

[0090] In this embodiment, the hyperbolic tangent function is used to obtain the normalized value of the abnormal score of the discrete coefficient of non-boundary points. The hyperbolic tangent function is a well-known technique and will not be described in detail in this application.

[0091] All suspected noise points on the z-th structural surface are classified according to their reflection intensity. The dispersion of the modulus of all suspected noise points in each category is denoted as the color dispersion. The curvature of each suspected noise point is calculated, and the dispersion of the curvature of all suspected noise points in each category is denoted as the curvature dispersion. The mean of the color dispersion and the mean of the curvature dispersion in each category is used as the discrete mean of that category. The discrete mean reflects whether the color and curvature of all suspected noise points in each category are consistent when the reflection intensity is consistent, thus reflecting whether all suspected noise points in each category belong to the same region. The larger the discrete mean, the greater the difference among all suspected noise points in each category, and the greater the probability that a suspected noise point in each category is a noise point. The calculation of curvature is a known technique and will not be described in detail in this application.

[0092] In this embodiment, the k-means algorithm is used to classify all suspected noise points on the z-th structural surface according to their reflection intensity. The number of classes is obtained by the elbow rule, and the distance is the absolute value of the difference between the reflection intensities of the suspected noise points. The k-means algorithm is a well-known technology and will not be described in detail here. As other implementation methods, based on the ability to classify all suspected noise points on the z-th structural surface according to their reflection intensity, implementers may use other existing technologies, such as the Density Peak Clustering Algorithm (DPC), etc. This application does not impose any special restrictions.

[0093] In this embodiment, the dispersion of the modulus and the dispersion of the curvature are both variances. As other implementation methods, based on the ability to measure the unevenness of the distribution of the modulus and the unevenness of the distribution of the curvature, the implementer may use other existing technologies, such as standard deviation, coefficient of variation, etc. This application does not impose any special restrictions.

[0094] Following the method for obtaining the discrete mean of each class in the z-th structural surface, obtain the discrete mean of each class in the remaining structural surfaces.

[0095] Furthermore, in irregularly shaped buildings, although key areas often exhibit certain geometric changes, these changes typically possess spatial continuity and concentration characteristics.

[0096] Based on the above analysis, a region growing method is used to obtain growth regions for each structural surface based on the normal vector similarity between suspected noise points. The proportion of suspected noise points outside the growth regions is then counted among all suspected noise points. This proportion reflects the percentage of isolated points not belonging to any region among the suspected noise points on the inner surface. A larger proportion indicates more isolated points and greater noise interference on the z-th structural surface. The process of obtaining growth regions based on the normal vector similarity between suspected noise points using the region growing method is a well-known technique and will not be elaborated upon here.

[0097] Furthermore, by combining the discrete mean values ​​of each type of structural surface in each building region with the aforementioned proportion of each structural surface in each building region, the point cloud distortion coefficient of each building region is obtained, expressed as:

[0098] In the formula, This represents the point cloud distortion coefficient for the u-th building region; This represents the average of the discrete means of all classes across all structural surfaces in the u-th building region; This represents the average percentage of the number of such structural surfaces in all structural surfaces of the u-th building region; This represents the reliability coefficient of the u-th building area.

[0099] It should be noted that the point cloud distortion coefficient reflects the degree of interference experienced by the u-th building area. A larger point cloud distortion coefficient indicates a greater deviation of the point cloud data from the true geometric structure of the u-th building area, necessitating more focused noise reduction to improve the accuracy of subsequent surveying and the scientific rigor of cost assessment. A schematic diagram illustrating the process of obtaining the point cloud distortion coefficient is shown below. Figure 3 As shown.

[0100] Step 5: Use a statistical filtering algorithm to denoise the point cloud data of each building area, and then combine it with the image to obtain a 3D model of the building to be surveyed.

[0101] Statistical filtering algorithms are used to denoise the point cloud data of each building region. During the denoising process, the negative of the point cloud distortion coefficients for each building region is used as the exponent of an exponential function with the natural constant as the base. The result of this exponential function is then used as the standard deviation multiple for denoising the point cloud data of each building region using the statistical filtering algorithm. A smaller truncation neighborhood is set when the interference level is higher to avoid losing important details due to global smoothing. The statistical filtering algorithm is a well-known technique and will not be described in detail here.

[0102] Import the point cloud model of the building to be surveyed, after removing non-building area point clouds and denoising, into BIM software such as Revit for 3D modeling. Export the resulting 3D model in .obj format. Import the .obj file into 3DS Max software, then load multi-angle high-resolution images of the building into the material editor within 3DS Max. Apply textures according to the correspondence between the 3D model and the images to obtain a textured 3D model. Finally, export the quantities of the 3D building model using 3DS Max software.

[0103] Based on the same inventive concept as the above method, this application also provides a high-precision building surveying system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described high-precision building surveying methods.

[0104] In summary, this application obtains a difference coefficient by measuring the spatial overlap, color difference, and reflectivity difference between point cloud regions. This quantifies the difference between each point cloud region and its surrounding areas, which helps in the subsequent identification of occluding objects that are not part of the building itself, thus improving the realism and usability of the point cloud data. By identifying building regions and removing point cloud data outside of these regions, the computational load of subsequent processing is reduced. Filling holes in the removed point cloud model ensures the integrity of the point cloud model.

[0105] Furthermore, by obtaining reliability coefficients, the degree of interference to the point cloud data of each building area can be assessed, providing dynamic parameter basis for subsequent denoising processing. By filtering suspected noise points using discrete coefficients and analyzing the color and curvature consistency of these points under the condition of consistent reflection intensity, combined with the spatial continuity and concentration characteristics of suspected noise points, point cloud distortion coefficients are obtained. This quantifies the actual degree of interference to the point cloud data of each building area, providing more accurate parameter basis for subsequent denoising processing. Dynamically adjusting the parameters in the statistical filtering algorithm based on the point cloud distortion coefficients enables targeted denoising according to the noise characteristics of different areas, preserving key architectural details, improving the accuracy of the 3D model, and combining images to apply textures to the 3D model, generating a textured 3D model. This provides more accurate and intuitive data support for subsequent architectural surveying and engineering cost assessment.

[0106] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0107] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from its essential characteristics. Therefore, the embodiments described above should be considered exemplary and non-limiting in all respects.

Claims

1. A high-precision building surveying method, characterized in that, The method includes the following steps: Collect point cloud models and images of the building to be surveyed. The point cloud model contains the color and reflection intensity of the point cloud data. The point cloud model is divided into point cloud regions by point cloud segmentation method. The difference coefficient of each point cloud region is obtained by the spatial overlap, color difference and reflection intensity difference of point cloud data between each point cloud region and other point cloud regions. Then, each building region is identified and the point cloud data outside the building region is removed from the point cloud model. The hole is filled in the removed point cloud model. The structural surfaces of each building area are segmented to obtain the structural surfaces and their interior point ratios. Boundary points and non-boundary points of each structural surface are also obtained. The dispersion of reflection intensity of boundary points and non-boundary points within a preset nearest neighbor range is used to obtain the dispersion coefficients of each boundary point and non-boundary point. This allows for the selection of suspected noise points from all non-boundary points on each structural surface. The reliability coefficient of each building area is obtained by combining the interior point ratio of each structural surface with the distribution of the dispersion coefficients of all boundary points. Suspected noise points in each structural surface are classified according to reflection intensity. The point cloud distortion coefficient of each building area is obtained by combining the color and curvature dispersion of suspected noise points in each category within each building area, as well as the normal vector similarity between suspected noise points in each structural surface, with the reliability coefficients. A statistical filtering algorithm is then used to denoise the point cloud data of each building area. Finally, combined with the image, a 3D model of the building to be surveyed is obtained. The process of obtaining the difference coefficient is as follows: Obtain the AABB bounding box of each point cloud region, and obtain the overlap value of the AABB bounding box between each point cloud region and each other point cloud region on the three coordinate axes; denot each point cloud region whose overlap value with each other on the three coordinate axes is positive as each connected point cloud region of each point cloud region. Calculate the magnitude of the vector composed of the R, G, and B color components of each point cloud data in each point cloud region; calculate the arithmetic mean of the magnitudes of all point cloud data in each point cloud region; and calculate the average value of the reflection intensity of all point cloud data in each point cloud region. The average of the differences between the arithmetic mean of each point cloud region and all its connected point cloud regions is denoted as the first mean; the average of the differences between the average of each point cloud region and all its connected point cloud regions is denoted as the second mean. The difference coefficients are directly proportional to the first mean and the second mean, respectively, and inversely proportional to the number of connected point cloud regions in each point cloud region. The method for obtaining the discrete coefficients is as follows: The ratio of the dispersion of the reflection intensity of all boundary points within the preset nearest neighbor range of each boundary point of each structural surface to the number of boundary points is used as the dispersion coefficient of each boundary point of each structural surface. The ratio of the dispersion of the reflection intensity of all non-boundary points within the preset nearest neighbor range of each non-boundary point of each structural surface to the number of non-boundary points is used as the dispersion coefficient of each non-boundary point of each structural surface. The method for obtaining the reliability coefficient is as follows: An anomaly detection algorithm is used to obtain the anomaly score of the discrete coefficient of each boundary point of each structural surface, and the proportion of boundary points with normalized anomaly scores greater than a preset threshold in each structural surface is counted. The reliability coefficient is directly proportional to the average of the interior point ratio of all structural surfaces of each building surface, and inversely proportional to the average of the proportion of the number of all structural surfaces of each building surface. The process of obtaining the point cloud distortion coefficient is as follows: Calculate the normal vector of each suspected noise point, and use the region growing method to obtain each growing region in each structural surface based on the similarity of the normal vectors between suspected noise points in each structural surface. Calculate the proportion of suspected noise points outside the growing region in each structural surface among all suspected noise points. The dispersion of the modulus of all suspected noise points in each class is denoted as the color dispersion; the mean of the color dispersion and the curvature dispersion in each class is taken as the discrete mean of each class. The expression for the point cloud distortion coefficient of each building area is as follows: In the formula, This represents the point cloud distortion coefficient for the u-th building region; This represents the average of the discrete means of all classes across all structural surfaces in the u-th building region; This represents the average percentage of the number of such structural surfaces in all structural surfaces of the u-th building region; This represents the reliability coefficient of the u-th building area.

2. The high-precision building surveying method as described in claim 1, characterized in that, The calculation process of the difference coefficient is as follows: calculate the sum of the first mean and the second mean, map the number of connected point cloud regions to a positive number, and the difference coefficient is the ratio of the sum to the positive number.

3. The high-precision building surveying method as described in claim 1, characterized in that, The method for identifying each building area and removing point cloud data outside the building areas from the point cloud model is as follows: Obtain the segmentation threshold of the difference coefficient of all point cloud regions, and regard the point cloud regions with the difference coefficient less than or equal to the segmentation threshold as building regions, and remove the point cloud data outside the building regions from the point cloud model.

4. The high-precision building surveying method as described in claim 1, characterized in that, The process for obtaining the suspected noise points is as follows: An anomaly detection algorithm is used to obtain the anomaly score of the discrete coefficients of each non-boundary point of each structural surface. Boundary points whose normalized anomaly score is greater than a preset threshold are identified as suspected noise points.

5. The high-precision building surveying method as described in claim 1, characterized in that, In the process of using the statistical filtering algorithm to denoise the point cloud data of each building area, the negative of the point cloud distortion coefficient of each building area is used as the exponent of an exponential function with the natural constant as the base, and the calculation result of the exponential function is used as the standard deviation multiple when the statistical filtering algorithm denoises the point cloud data of each building area.

6. A high-precision building surveying system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the high-precision building surveying method as described in any one of claims 1-5.

Images